MoinMoin: MoinMoinChat/Logs/moin-dev/2008-08-15

2008-08-15T00:17:31  <dennda> umm
2008-08-15T03:07:03  <CIA-54> Pawel Pacana <pawel.pacana@gmail.com> default * 4807:ef0351b8b5a2 1.8-mercurialbackend-ppacana/MoinMoin/storage/backends/hg.py: (hg) Making performance suck less - _list_revision caching.
2008-08-15T03:07:03  <CIA-54> Pawel Pacana <pawel.pacana@gmail.com> default * 4808:dc0188702746 1.8-mercurialbackend-ppacana/MoinMoin/storage/backends/hg.py: (hg) Unused imports purge.
2008-08-15T03:10:52  <PawelPacana> at least conversion gained some decent speed :o
2008-08-15T03:11:48  <PawelPacana> few seconds faster than FSBackend on converting underlay
2008-08-15T03:17:17  <dennda> heh
2008-08-15T03:17:35  <dennda> You may only be satisfied if you outperform the MemoryBackend!
2008-08-15T03:17:42  <dennda> ;)
2008-08-15T04:11:38  <dennda> Wow!
2008-08-15T04:11:52  * dennda just ran a wiki *with* underlay and *without* conversion
2008-08-15T04:12:31  <dennda> Using the nifty new wrapper backend I just wrote. Didn't think it worked on first try, but it either did or I am too tired to see any mistake I made
2008-08-15T04:12:54  <dennda> just cool. more on that tomorrow
2008-08-15T04:21:58  <dennda> test_backends_wrapper.py[111] ...............................................................................................................
2008-08-15T04:22:05  <dennda> tests finished: 111 passed in 0.16 seconds
2008-08-15T04:22:07  <dennda> ;)
2008-08-15T08:01:09  <dreimark> moin
2008-08-15T08:45:00  <ThomasWaldmann>   moin
2008-08-15T10:28:09  <zenhase> moin
2008-08-15T10:30:40  <CIA-54> Florian Krupicka <florian.krupicka@googlemail.com> default * 4082:748d0cbcbd9b 1.8-wsgi-fkrupicka/MoinMoin/web/request.py: Fix: variable assignment in evaluate_request
2008-08-15T10:30:40  <CIA-54> Florian Krupicka <florian.krupicka@googlemail.com> default * 4083:bafefd635670 1.8-wsgi-fkrupicka/MoinMoin/web/contexts.py: Fix: typos introduced while porting check_spider
2008-08-15T11:13:49  <dreimark> Beating the (sh** out of the) GIL - Multithreading vs. Multiprocessing (Guy K. Kloss)
2008-08-15T11:13:58  <dreimark> https://wiki.sistec.dlr.de/TechTalks/20080813KlossMultithreadingVsMultiprocessing
2008-08-15T11:16:43  <xorAxAx> https://wiki.sistec.dlr.de/TechTalks/20080730WendelCsharp#head-6d6a6f3500f3d6a7d53176343451d91bc63ff067
2008-08-15T11:16:52  <xorAxAx> he left the realm of python
2008-08-15T11:32:21  <zenhase> good ol' GIL
2008-08-15T11:57:33  <dennda> Hmpf. Why did my server go down?
2008-08-15T12:19:22  <ThomasWaldmann> gizmo__: where are your cleanup results?
2008-08-15T12:42:13  <dennda> lol
2008-08-15T12:42:38  <dennda> "Yes, the system shut down and rebooted this morning. All other VPS as well. We don't know why."
2008-08-15T12:44:03  <ThomasWaldmann> vermutlich wegen erhoehtem neutrinoaufkommen :P
2008-08-15T12:45:33  <xorAxAx> dennda: which provider?
2008-08-15T12:48:29  <dennda> hosteurope
2008-08-15T13:01:57  <CIA-54> Thomas Waldmann <tw AT waldmann-edv DOT de> default * 3788:d3d1b3f9f159 1.7/docs/CHANGES: indexing filters: added a note to CHANGES about the previous changeset and its implications
2008-08-15T15:21:35  <ThomasWaldmann> re
2008-08-15T15:24:53  <CIA-54> Christopher Denter <moin GUESSWHAT the DASH space DASH station ROUNDTHING com> default * 4807:4027e04fb26b 1.8-storage-cdenter/MoinMoin/storage/ (_tests/test_backends_wrapper.py backends/wrapper.py): storage: add first naive implementation of a wrapper backend that takes two backends, one of which readonly
2008-08-15T15:24:55  <CIA-54> Christopher Denter <moin GUESSWHAT the DASH space DASH station ROUNDTHING com> default * 4808:63d439ad3a63 1.8-storage-cdenter/MoinMoin/PageEditor.py: storage: PageEditor: Allow creation of pages that have been deleted earlier (Thanks JohannesBerg)
2008-08-15T15:26:06  <dennda> http://paste.pocoo.org/show/82280/ <-- Example wikiconfig_local.py for that wrapper thing
2008-08-15T15:26:13  <dennda> xorAxAx: Did you look at the failing bug?
2008-08-15T15:26:18  <dennda> s/bug/test/
2008-08-15T15:57:14  <dreimark> ROWrapperBackend sounds somehow strange (because row is a legal word)
2008-08-15T16:01:41  <xorAxAx> dennda: oh, indeed
2008-08-15T16:02:03  <ThomasWaldmann> dennda: i guess you rather mean "directed" than "deferred"
2008-08-15T16:02:07  <xorAxAx> dennda: i posed you a question in return
2008-08-15T16:02:15  <xorAxAx> dennda: still waiting for the answer :)
2008-08-15T16:05:04  <ThomasWaldmann> dennda: maybe the wrapper docstring should also mention that of course also a 2nd new backend with the sys/help pages can be used as "underlay", not only the fs17 one
2008-08-15T16:05:40  <ThomasWaldmann> +# Problems: What happens if an Item of the second_backend is renamed? Think deeper about such problems and add behaviour where missing
2008-08-15T16:05:55  <ThomasWaldmann> that doesn't make sense as the 2nd backend is the readonly one
2008-08-15T16:07:28  <ThomasWaldmann> and maybe that thing should be called something like "layering backend"
2008-08-15T16:08:02  <ThomasWaldmann> rather than ROW rapper :P
2008-08-15T16:09:34  <ThomasWaldmann> + for item in self.first.iteritems():
2008-08-15T16:09:35  <ThomasWaldmann> + items.append(item)
2008-08-15T16:09:42  <ThomasWaldmann> this will kill the fs backend
2008-08-15T16:10:14  <xorAxAx> well, thats a bug in the fs backend probably :)
2008-08-15T16:10:52  <ThomasWaldmann> probably, but as we know that and as it is the only working backend with no special dependencies we have, we should avoid that
2008-08-15T16:11:35  <ThomasWaldmann> esp. if it is avoided easily as in that case
2008-08-15T16:12:15  <ThomasWaldmann> dennda: same problem in history() btw
2008-08-15T16:13:50  <ThomasWaldmann> btw, later we can generalize that idea not to hardcode first and second, but to use a list
2008-08-15T16:16:14  <xorAxAx> generally its a good idea to avoid large lists of items, yes
2008-08-15T16:17:54  <dreimark> bbl
2008-08-15T16:18:08  <ThomasWaldmann> dennda: many (all?) docstrings miss epydoc stuff like @param @returns etc. and are often unclear about object types
2008-08-15T16:19:57  <ThomasWaldmann> also, i am not sure if yielding item revs in history is a good idea
2008-08-15T16:20:23  <ThomasWaldmann> (one could also yield those tuples)
2008-08-15T16:21:23  <ThomasWaldmann> wrapper.history is creating a list of ALL revision objects -> that for sure kills fs backend
2008-08-15T16:22:04  <ThomasWaldmann> and that I have to comment on that basically means that it was not practically tested
2008-08-15T16:25:07  <ThomasWaldmann> dennda: ?
2008-08-15T16:50:40  <waldi> hmm, I broke it
2008-08-15T16:50:41  <waldi> 2008-08-15 16:50:31,307 INFO MoinMoin.Page:1344 failed to load "HelpOnMoinWikiSyntax" cache: 'Element' object has no attribute 'copy'
2008-08-15T17:08:52  <xorAxAx> dennda: waldmannish communication pattern? ask and vanish :)
2008-08-15T17:31:16  <dennda> sorry, was off for lunch with friends. just returned
2008-08-15T17:32:10  <dennda> xorAxAx: I did answer that question, didn't I?
2008-08-15T17:32:33  <dennda> dreimark: Yes, it's the ROW rapper straight from the bronx! :)
2008-08-15T17:32:36  <xorAxAx> dennda: indeed, but i wasnt satisfied :)
2008-08-15T17:32:44  <dennda> xorAxAx: how so?
2008-08-15T17:32:44  <xorAxAx> dennda: ok, let me clarify:
2008-08-15T17:32:58  <xorAxAx> dennda: i was wondering whether it is good style to introduce kwargs into the api
2008-08-15T17:33:14  <xorAxAx> but i think we can do so
2008-08-15T17:33:35  <dennda> I don't know why it'd be a bad idea
2008-08-15T17:34:10  <xorAxAx> i read people disliking kwargs on LtU IIRC, but i didnt understand them either :)
2008-08-15T17:34:23  <dennda> LtU?
2008-08-15T17:35:07  <xorAxAx> lambda the ultimate
2008-08-15T17:35:17  <xorAxAx> good community driven blog
2008-08-15T17:35:46  <dennda> never heard of it
2008-08-15T17:37:14  <dennda> ThomasWaldmann: I am aware of all the problems you mentioned. I planned to rephrase the docstrings to be compliant to epydocs rules
2008-08-15T17:37:24  <dennda> (which I already mentioned but got terribly misunderstood) :)
2008-08-15T17:37:53  <dennda> xorAxAx: as neither of us has any real arguments against a single kwarg, I think we should just keep it
2008-08-15T17:39:58  <dennda> ThomasWaldmann: You may, however, want to edit a page that only exists in the read-only backend, in which case you'd need to copy it to the first backend
2008-08-15T17:40:07  <dennda> (and if it gets deleted, do a whiteout or something)
2008-08-15T17:42:06  <dennda> or is that an invalid usecase?
2008-08-15T17:42:14  <dennda> in that case I'd have to do less work
2008-08-15T17:43:06  <CIA-54> Alexander Schremmer <alex AT alexanderweb DOT de> default * 4809:36c9c9698a76 1.8-storage-cdenter/MoinMoin/storage/ (3 files in 2 dirs): Add exception handling to the tracing backend, allow kwargs.
2008-08-15T17:43:07  <CIA-54> Alexander Schremmer <alex AT alexanderweb DOT de> default * 4810:b81225b4723f 1.8-storage-cdenter/MoinMoin/ (3 files in 3 dirs): Merge storage repo.
2008-08-15T17:47:14  <dennda> thanks
2008-08-15T17:49:16  <dennda> Umm
2008-08-15T17:49:39  <dennda> Why are there so few sourcecode tests as of late? There are only 49 now. I remember there were much more once
2008-08-15T17:49:59  <CIA-54> Christopher Denter <moin GUESSWHAT the DASH space DASH station ROUNDTHING com> default * 4811:8877954807ba 1.8-storage-cdenter/MoinMoin/storage/_tests/test_backends.py: storage: pep8 fix
2008-08-15T17:52:10  * dennda goes and buys water
2008-08-15T17:54:01  <xorAxAx> dennda: they only test "new" sourcefiles. not sure if that is completly sound, but at least it reduces the amount of files sometimes
2008-08-15T17:59:34  <dennda> ok if that is on purpose it's fine with me. just thought I destroyed something that'd backfire when you ran it as of the evaluation process :)
2008-08-15T18:00:31  <ThomasWaldmann> if you like to trigger a full sourcecode scan, just touch everything
2008-08-15T18:01:48  <dennda> ok thanks
2008-08-15T19:15:49  <ThomasWaldmann> dennda: gizmo__: PawelPacana: waldi: zenhase: please merge moin/1.8 (0 failing tests there, including a full test_sourcecode scan)
2008-08-15T19:15:58  <ThomasWaldmann> ====================================================== tests finished: 1164 passed, 67 skipped in 134.10 seconds =======================================================
2008-08-15T19:16:07  <dennda> oh
2008-08-15T19:16:15  <dennda> I was just fixing some tests in my branch
2008-08-15T19:16:23  <dennda> shall I merge first?
2008-08-15T19:16:51  <ThomasWaldmann> dennda: i didn't fix branch specific tests, only the stuff that is in main branch
2008-08-15T19:18:13  <ThomasWaldmann> if you fixed same, maybe create a diff with all your current work, hg revert --all, pull from main branch, merge, and then reapply all hunks that are missing
2008-08-15T19:19:10  <dennda> no I didn't change any of those filter things
2008-08-15T19:19:37  <ThomasWaldmann> filter things?
2008-08-15T19:19:45  <dennda> MoinMoin/filter/... was modified
2008-08-15T19:19:56  <ThomasWaldmann> that's a win32 bugfix
2008-08-15T19:20:17  <ThomasWaldmann> but neutral to test outcome
2008-08-15T19:20:43  <dennda> just merged
2008-08-15T19:21:16  <dennda> there were 5 changesets that I merged
2008-08-15T19:40:50  * ThomasWaldmann runs tests in waldi's repo
2008-08-15T19:44:31  <dennda> just to get that right: PageEditor.copyPage(newpage) is meant to copy the current page to newpage. if newpage already exists, it should create a new revision for the copy. correct?
2008-08-15T19:46:28  <xorAxAx> dennda: i would suggest to fold that method into the caller
2008-08-15T19:46:35  <xorAxAx> hoping that there is only one
2008-08-15T19:46:43  <dennda> "the caller" being "the test"
2008-08-15T19:46:47  <dennda> :)
2008-08-15T19:46:49  <xorAxAx> nobody else?
2008-08-15T19:48:36  <dennda> dunno for sure. it's just the test complaining. lemme comment that out and copy a page :)
2008-08-15T19:49:21  <xorAxAx> dennda: my point is that the copy action should implement it, probably calling code that was factored out of the migration function
2008-08-15T19:50:07  <xorAxAx> i am suggesting this because it will probably make it easier
2008-08-15T19:50:24  <dennda> yes, let's see who complains
2008-08-15T19:50:37  <ThomasWaldmann> dreimark: could you please de-duplicate the common code in MoinMoin/macro/_tests/* (maybe to a common macro test base class?)?
2008-08-15T19:51:30  <ThomasWaldmann> dreimark: can be carefully done in 1.7, then pulled/merged into 1.8
2008-08-15T19:51:31  <dennda> ah
2008-08-15T19:51:40  <dennda> copypage was one of the things you need to allow in your config
2008-08-15T19:53:04  <ThomasWaldmann> dreimark: reason is that this code is failing in moin/1.8-dom-bblank and it should not be needed to fix the same code at 8 places...
2008-08-15T19:54:48  <dennda> xorAxAx: it's only used in a bunch of tests and the copypage action
2008-08-15T19:54:54  <dennda> as far as grep is concerned
2008-08-15T19:57:32  <xorAxAx> yeah, so i suggest to copy the code there
2008-08-15T19:57:37  <xorAxAx> if it makes it easier
2008-08-15T19:57:45  <xorAxAx> move the code i mean
2008-08-15T19:58:08  <dennda> sure
2008-08-15T19:58:10  <dennda> doing that
2008-08-15T19:59:08  <dennda> xorAxAx: but as for my originial question, the code is *meant* to do what I described, correct?
2008-08-15T19:59:26  <xorAxAx> dennda: no idea, ask reimar. for me that sounds like a bad semantics
2008-08-15T19:59:47  <dennda> ok
2008-08-15T20:00:07  <xorAxAx> you probably want overwrite (new revision) semantics, but only if the user agreed IMHO
2008-08-15T20:00:16  <xorAxAx> (or if its deleted)
2008-08-15T20:05:01  <dennda> which would need new gui elements, at least the former
2008-08-15T20:05:04  <dennda> bbiab
2008-08-15T20:27:00  <dennda> ah as far as I can tell, copyPage is just renaming but keeping the thing that was renamed
2008-08-15T20:41:13  <dennda> AttributeError: 'Request' object has no attribute 'data_backend'
2008-08-15T20:41:15  <dennda> ?
2008-08-15T20:51:49  <ThomasWaldmann> what's the question?
2008-08-15T20:52:32  <dennda> I was under the impression that request objects always had a reference to the data_backend. but I have no knowledge when it comes to the request-stuff
2008-08-15T20:52:51  <xorAxAx> hopefully they never have :)
2008-08-15T20:53:14  <dennda> they do have
2008-08-15T20:55:36  <xorAxAx> show me a single point where this is the case
2008-08-15T20:55:48  <xorAxAx> and i will show you some escher stairs in reality :)
2008-08-15T20:59:36  <dennda> stop asking catchy questions then
2008-08-15T20:59:48  <dennda> there are quite some places where backend = request.data_backend is used
2008-08-15T21:00:29  <xorAxAx> catchy questions?
2008-08-15T21:00:34  <xorAxAx> where are those place?
2008-08-15T21:00:35  <xorAxAx> s
2008-08-15T21:01:03  * xorAxAx didnt pose a question :)
2008-08-15T21:01:59  <dennda> request.cfg.data_backend, I mean
2008-08-15T21:02:04  <dennda> err
2008-08-15T21:02:05  <dennda> !
2008-08-15T21:09:04  * ThomasWaldmann hopes dennda has a todo item of "remove mixing/splitting meta/data in Page"
2008-08-15T21:09:40  <dennda> I am not sure what you mean. Which line?
2008-08-15T21:11:29  <ThomasWaldmann> all over the place :)
2008-08-15T21:11:47  <dennda> e.g.?
2008-08-15T21:11:56  <ThomasWaldmann> wikiutil.add_metadata_to_body
2008-08-15T21:12:49  <ThomasWaldmann> backend has separate md/d, that ^^ mixes them together, at some other place they are split/parsed again
2008-08-15T21:13:49  * dennda adds an appropriate todo item
2008-08-15T21:13:56  <dennda> of what importance do you think that is?
2008-08-15T21:14:04  <ThomasWaldmann> btw, with "todo" i don't mean now, but after soc
2008-08-15T21:14:11  <dennda> ah
2008-08-15T21:14:22  <dennda> you should explicitly say that :)
2008-08-15T21:14:34  <ThomasWaldmann> i did :)
2008-08-15T21:14:43  <dennda> btw, I got at least one failing test due to ACLs missing
2008-08-15T21:14:47  <dennda> what to do about that one?
2008-08-15T21:15:18  <ThomasWaldmann> is it in your pushed code?
2008-08-15T21:16:13  <dennda> the failing test?
2008-08-15T21:16:24  <ThomasWaldmann> yes
2008-08-15T21:16:29  <dennda> it's failing since quite some time. I am just working on getting the tests work again
2008-08-15T21:16:44  <dennda> just finished one, but this one obviously cannot work
2008-08-15T21:16:55  * ThomasWaldmann runs tests
2008-08-15T21:16:58  <dennda> test_copy_page_acl_no_read in test_PageEditor
2008-08-15T21:20:15  <ThomasWaldmann> >           ItemAlreadyExistsError: An Item with the name u'AutoCreatedMoinMoinTemporaryTestPageX' already exists!
2008-08-15T21:20:32  <dennda> yes, sure it does
2008-08-15T21:20:47  <dennda> the tests there are fixed locally. didn't push yet
2008-08-15T21:20:55  <dennda> but I cannot fix that ACL thing
2008-08-15T21:21:04  <ThomasWaldmann> commit/push
2008-08-15T21:21:19  <dennda> ok just a sec
2008-08-15T21:24:51  <xorAxAx> sigh, pushing failing tests is bd
2008-08-15T21:25:00  <xorAxAx> skipping them with an appropriate note is much better
2008-08-15T21:25:16  <dennda> actually the test failed before already. the next push will have less failing tests than the current one
2008-08-15T21:25:31  <dennda> yes, skipping may be what I need to do here
2008-08-15T21:28:30  <CIA-54> Christopher Denter <moin GUESSWHAT the DASH space DASH station ROUNDTHING com> default * 4818:e63007a625e0 1.8-storage-cdenter/MoinMoin/ (PageEditor.py _tests/test_PageEditor.py): storage: Add missing behavior to copyPage and fix tests accordingly.
2008-08-15T21:28:41  <xorAxAx> (also commiting failing tests is bad, but most people dont have a workflow that is dvcs affine enough to avoid that)
2008-08-15T21:28:43  <dennda> ThomasWaldmann: ^^^
2008-08-15T21:29:03  * ThomasWaldmann runs tests
2008-08-15T21:30:09  * dennda informs himself how to skip tests with py.test
2008-08-15T21:31:49  <xorAxAx> use something that describes the test failure and some goal how important it is to fix the test and when you plan to fix it :)
2008-08-15T21:35:10  <dennda> "ACLs not integrated with storage yet. Fix after SoC."?
2008-08-15T21:35:18  <ThomasWaldmann> dennda: those acl tests are complete nonsense now
2008-08-15T21:36:11  <dennda> ThomasWaldmann: just remove the last? (test_copy_page_acl_read is still useful if it is renamed to test_copy_page_to_existing_page)
2008-08-15T21:37:40  <ThomasWaldmann> you have to fix them so they test what they did before
2008-08-15T21:38:16  <dennda> going to be a bit hard without acl support
2008-08-15T21:38:50  <dennda> "complete nonsense" as of the new storage not yet supporting acls or after I changed them?
2008-08-15T21:39:12  <xorAxAx> well, whats your acl-idea, dennda?
2008-08-15T21:39:19  <ThomasWaldmann> what do you mean by "without acl support"?
2008-08-15T21:40:06  <dennda> i don't have any right now. johill told me to "just forget about acls completely"
2008-08-15T21:40:10  <dennda> as of SoC
2008-08-15T21:40:39  <xorAxAx> dennda: ok, fair enough
2008-08-15T21:41:51  <dennda> ThomasWaldmann: well, the last test there indicates that the current state of the software does not care whatsoever about the acl defined there, which you can see by looking at the assert statement. the page gets copied. (which is indicated by revision being 3, which is != 2
2008-08-15T21:44:45  <ThomasWaldmann> dennda: that acl line is never used
2008-08-15T21:44:52  <dennda> exactly!
2008-08-15T21:45:09  <dennda> that's what I am talking of :)
2008-08-15T21:45:14  <xorAxAx> which acl line?
2008-08-15T21:45:14  <ThomasWaldmann> no
2008-08-15T21:45:29  <dennda> that's what I thought I was talking of :)
2008-08-15T21:45:32  <xorAxAx> the storage backend shouldnt read acl lines of the wiki page content :)
2008-08-15T21:45:32  <ThomasWaldmann> it is never used because it never gets used by the test code
2008-08-15T21:45:39  <dennda> xorAxAx: of course
2008-08-15T21:46:51  <xorAxAx> dennda: and as long there is no defined acl storage format, one cant formulate a test for it :)
2008-08-15T21:47:00  * dennda skips it
2008-08-15T21:47:51  <dennda> and copy-renames that other test to have an additional test for something that wasn't tested yet
2008-08-15T21:48:02  <ThomasWaldmann> even if you had acl code doing something, those tests would not work
2008-08-15T21:48:11  <dennda> yes, just saw that
2008-08-15T21:48:16  <ThomasWaldmann> because self.text is never used
2008-08-15T21:48:21  <dennda> yes
2008-08-15T21:48:35  <ThomasWaldmann> and if you try createPage repeatedly, it will fail
2008-08-15T21:48:43  <dennda> yes
2008-08-15T21:48:49  <ThomasWaldmann> so it's all fscked up :P
2008-08-15T21:49:01  <dennda> just wait a sec
2008-08-15T21:49:24  <ThomasWaldmann> (just because someone thought we never need a remove_item thing)
2008-08-15T21:49:40  <xorAxAx> ThomasWaldmann: well, we talked a lot about it
2008-08-15T21:50:18  <ThomasWaldmann> how's that related to a still existing problem? :)
2008-08-15T21:50:27  <xorAxAx> which problem?
2008-08-15T21:50:50  <ThomasWaldmann> that you can not remove an item, not even for testing
2008-08-15T21:51:05  <xorAxAx> why would you need to that in testing?
2008-08-15T21:51:32  <ThomasWaldmann> see the code we talk about since 20 mins
2008-08-15T21:51:51  <xorAxAx> i guess that wont answer my quetion :)
2008-08-15T21:52:03  <ThomasWaldmann> sure it does
2008-08-15T21:52:31  <ThomasWaldmann> the stuff that is commented out just can't work just because of that
2008-08-15T21:52:48  <ThomasWaldmann> of course we can invent funny new pagenames for every test :)
2008-08-15T21:53:09  <xorAxAx> umm
2008-08-15T21:53:40  <xorAxAx> well, it cant demand a new item because that would a storage violation
2008-08-15T21:53:46  <xorAxAx> hmm, abstract violation
2008-08-15T21:54:05  <xorAxAx> just nuking pages for testing purposes seems like a "hack" to me as well
2008-08-15T21:54:27  <xorAxAx> so probably the best idea is to give every function a new context
2008-08-15T21:54:54  <xorAxAx> or provide a "needs context reset" operation
2008-08-15T21:55:06  <xorAxAx> that means "i cant clean up conceptually" :)
2008-08-15T21:55:17  * ThomasWaldmann estimates tests being at least 10x slower after storage/wsgi merge
2008-08-15T21:55:21  <CIA-54> Christopher Denter <moin GUESSWHAT the DASH space DASH station ROUNDTHING com> default * 4819:9392d255c662 1.8-storage-cdenter/MoinMoin/ (_tests/test_PageEditor.py config/multiconfig.py): storage: tests: skipping ACL tests for now. Use them when the needed functionality has been implemented. Add new test for copyPage
2008-08-15T21:55:37  <xorAxAx> ThomasWaldmann: huh? the storage branch should run them much faster
2008-08-15T21:55:49  <xorAxAx> because it doesnt/shouldnt use the full underlay
2008-08-15T21:55:54  <dennda> ...and your harddrive will be used less
2008-08-15T21:56:13  <ThomasWaldmann> i dont have a hard disk :P
2008-08-15T21:56:15  <xorAxAx> yeah, everything will have been become much better
2008-08-15T21:56:25  <dennda> which makes baby jesus smile
2008-08-15T22:00:25  <dennda> ThomasWaldmann: flash disk? even worse!
2008-08-15T22:00:36  <ThomasWaldmann> 2,5mins in 1.8
2008-08-15T22:01:53  <xorAxAx> dont forget to subtract the n-1 backends to avoid bragging with waldi-class benchmark figures :)
2008-08-15T22:01:53  * dennda benchmarks
2008-08-15T22:02:25  <dennda> still, 1.8 is much slower
2008-08-15T22:03:06  <ThomasWaldmann> 1,9 mins in 1.8-storage
2008-08-15T22:03:25  <ThomasWaldmann> (but quite some more tests :)
2008-08-15T22:03:44  <dennda> http://paste.pocoo.org/show/82329/
2008-08-15T22:03:53  <ThomasWaldmann> ok, so dennda is allowed to make it a bit slower :)
2008-08-15T22:04:45  <dennda> double amount of tests, half the time to run the tests :p
2008-08-15T22:04:48  <dennda> (roughly)
2008-08-15T22:05:58  <ThomasWaldmann> ok, so how you want to solve the problem?
2008-08-15T22:06:06  <dennda> which problem?
2008-08-15T22:06:25  <ThomasWaldmann> that it is not possible to remove items
2008-08-15T22:06:55  <ThomasWaldmann> and one maybe does not want to invent funny new names all the time just because of that
2008-08-15T22:07:00  <xorAxAx> yes
2008-08-15T22:07:12  <dennda> I guess I am not going to change that before SoC ends
2008-08-15T22:07:19  <xorAxAx> i can try to fix that
2008-08-15T22:07:28  <xorAxAx> the whole test running logic
2008-08-15T22:07:36  <xorAxAx> shouldnt be hard
2008-08-15T22:07:56  <ThomasWaldmann> new clean backend per test class?
2008-08-15T22:08:06  <xorAxAx> no
2008-08-15T22:08:25  <ThomasWaldmann> wouldn't help for tests structured like that one
2008-08-15T22:08:27  <xorAxAx> dennda: is your repo rather new wrt main repo?
2008-08-15T22:08:30  <xorAxAx> ThomasWaldmann: i know :)
2008-08-15T22:08:37  <xorAxAx> s/new/fresh/ :)
2008-08-15T22:08:39  <dennda> xorAxAx: I merged 1.8 a few hours ago
2008-08-15T22:08:57  <xorAxAx> dennda: neat
2008-08-15T22:09:07  <xorAxAx> ThomasWaldmann: why did you move the request instantiation down the "path" again`
2008-08-15T22:09:15  <xorAxAx> (what were the failing use cases)
2008-08-15T22:09:41  <ThomasWaldmann> ?
2008-08-15T22:10:00  <xorAxAx> 28442c70f757
2008-08-15T22:10:02  <xorAxAx> 5 days ago :)
2008-08-15T22:10:19  <dennda> hg blame rocks, doesn't it?
2008-08-15T22:10:23  <xorAxAx> hg log in this case
2008-08-15T22:10:32  <xorAxAx> i reviewed that change set
2008-08-15T22:10:35  <dennda> ah ok
2008-08-15T22:10:39  <xorAxAx> and just needed to look up the rev no :)
2008-08-15T22:11:03  <ThomasWaldmann> hunk 2 -> you don't need to create a request if you don't use it
2008-08-15T22:11:32  <xorAxAx> yes. but where did it fail with the module wide approah?
2008-08-15T22:11:33  <ThomasWaldmann> hunk 3 -> acl stuff was fscked before that change
2008-08-15T22:13:13  <xorAxAx> which acl stuff?
2008-08-15T22:13:25  <xorAxAx> some tests?
2008-08-15T22:14:46  <xorAxAx> then i will see them failing ...
2008-08-15T22:15:53  <ThomasWaldmann> the whole TestConfig this is cursed :)
2008-08-15T22:16:00  <ThomasWaldmann> s/this/thing/
2008-08-15T22:16:35  <ThomasWaldmann> has to get fixed in main branch, of course
2008-08-15T22:16:44  <xorAxAx> yes, but i am not sure whether thats the right fix for it
2008-08-15T22:17:05  <ThomasWaldmann> no, it is just some easy fix
2008-08-15T22:17:18  <ThomasWaldmann> the real fix should create the cfg in a sane way
2008-08-15T22:18:22  <xorAxAx> easy fix != workaround
2008-08-15T22:18:41  <xorAxAx> if the problem is the broken testconfig class, then just creating a request instance for every function is a workaround
2008-08-15T22:19:12  <xorAxAx> but we can probably live with that
2008-08-15T22:19:38  <ThomasWaldmann> call it like you want. it was long broken and currently works.
2008-08-15T22:20:23  <CIA-54> Alexander Schremmer <alex AT alexanderweb DOT de> default * 3943:25d4db484bf1 1.8/MoinMoin/_tests/test_sourcecode.py: Multiple extension modules are called xattr, check for the xattr function as well.
2008-08-15T22:20:27  <ThomasWaldmann> but there are some nasty side effects requiring additional code as you see in the acl tests
2008-08-15T22:20:37  <xorAxAx> it works is no really useful predicate
2008-08-15T22:20:47  <xorAxAx> which are "the acl tests"?
2008-08-15T22:20:50  <xorAxAx> (which modules)
2008-08-15T22:21:06  <ThomasWaldmann> just revert my change to see them failing
2008-08-15T22:21:13  <xorAxAx> ok
2008-08-15T22:22:30  <xorAxAx> hmm, is there nothing like reverse apply ...
2008-08-15T22:22:34  <xorAxAx> backout wants to commit
2008-08-15T22:23:18  <ThomasWaldmann> hg export + patch -p1
2008-08-15T22:23:24  <xorAxAx> | patch even, yep
2008-08-15T22:26:12  <xorAxAx> also, in other projects (that use TDD successfully), the tests need 11 hours to run completly
2008-08-15T22:26:33  <xorAxAx> the idea is to run tests selectively or use the distribution feature of py.test
2008-08-15T22:26:54  <xorAxAx> you can e.g. let the tests run on every core of the many servers that you have, ThomasWaldmann :)
2008-08-15T22:27:38  <xorAxAx> ThomasWaldmann: all tests pass for me
2008-08-15T22:27:54  <xorAxAx> with your changeset reverted
2008-08-15T22:28:49  <xorAxAx> (to use the distribution feature, one only needs to write a list of hostnames that are reachable by ssh and is mostly done)
2008-08-15T22:29:06  <xorAxAx> http://codespeak.net/py/dist/test.html#automated-distributed-testing
2008-08-15T22:29:18  <xorAxAx> ThomasWaldmann: so should i revert the changeset?
2008-08-15T22:29:36  <ThomasWaldmann> no
2008-08-15T22:29:49  <xorAxAx> then tell me how to reproduce the issue
2008-08-15T22:30:05  <xorAxAx> (the cs also contained another change that does "become ...")
2008-08-15T22:39:00  <ThomasWaldmann> xorAxAx: btw, the changeset has a rather long comment
2008-08-15T22:39:14  <xorAxAx> yes, but it doesnt answer my questions
2008-08-15T22:39:23  <xorAxAx> everything works without it ...
2008-08-15T22:40:13  <ThomasWaldmann> i try it right now
2008-08-15T22:41:34  <ThomasWaldmann> could be that the problem was just surfacing  in another repo
2008-08-15T22:45:30  <xorAxAx> dennda: hmm, the acl tests of pageeditor are failing because you ripped out the acl checks, not because it wouldnt work by design (i mean on this level, it ought to work :))
2008-08-15T22:45:52  <xorAxAx> dennda: but the idea to use copypage here is probably bad
2008-08-15T22:46:07  <xorAxAx> i mean, you skipped them mainly because of copypage, not because of acls :-)
2008-08-15T22:47:39  <xorAxAx> so i think i will clarify the skip
2008-08-15T22:48:33  <xorAxAx> dennda: i am getting a test failure
2008-08-15T22:48:38  <xorAxAx> that is related to packages
2008-08-15T22:49:00  <xorAxAx> i admit that the replaceunderlay command cannot be implemented with an appropriate replacement of the underlay stuff first
2008-08-15T22:49:14  <ThomasWaldmann> xorAxAx: the fix was for the storage repo.
2008-08-15T22:49:28  <xorAxAx> ThomasWaldmann: which ones fail?
2008-08-15T22:49:37  <xorAxAx> or should i find out ...
2008-08-15T22:49:43  <xorAxAx> (there is so much failing :))
2008-08-15T22:50:04  <ThomasWaldmann> try it. i merged it into storage 3 minutes after committing to 1.8 main
2008-08-15T22:52:00  <xorAxAx> omg, collectpackage
2008-08-15T22:52:41  <xorAxAx> ah
2008-08-15T22:52:56  <xorAxAx> dennda: i am seeing packagepages related test failures
2008-08-15T22:52:59  <xorAxAx> on your branch only
2008-08-15T22:58:04  <xorAxAx> omg, revision is 2
2008-08-15T23:01:25  <ThomasWaldmann> and by the magic of python, that even works :P
2008-08-15T23:02:04  <xorAxAx> no
2008-08-15T23:02:09  <xorAxAx> because of the magic of cpython
2008-08-15T23:03:55  <ThomasWaldmann> doesn't work in pypy?
2008-08-15T23:04:26  <xorAxAx> depends on your options
2008-08-15T23:04:40  <xorAxAx> (you can configure whether you want that optimization :)
2008-08-15T23:04:49  <xorAxAx> and other python implementations might behave differently as well
2008-08-15T23:11:37  <xorAxAx> ok, now i am seeing test failures that show broken root page behaviour
2008-08-15T23:16:01  <xorAxAx> ThomasWaldmann: i cant reproduce the issues either in the storage branch
2008-08-15T23:16:11  <xorAxAx> i suggest that we revert it and think about it again when it comes up
2008-08-15T23:16:17  <CIA-54> Alexander Schremmer <alex AT alexanderweb DOT de> default * 4820:b8157a53df9f 1.8-storage-cdenter/ (4 files in 3 dirs): Bind the backends lazily in the test wiki. Introduce a new decorator for tests that resets the backends, dont bind the backend early in RootPage.
2008-08-15T23:16:24  <CIA-54> Alexander Schremmer <alex AT alexanderweb DOT de> default * 4821:24d307333207 1.8-storage-cdenter/MoinMoin/_tests/test_PageEditor.py: Sanity fixes in the copy page tests, use the new decorator.
2008-08-15T23:16:25  <CIA-54> Alexander Schremmer <alex AT alexanderweb DOT de> default * 4822:526524632295 1.8-storage-cdenter/MoinMoin/_tests/test_packages.py: Dont lookup a page name on the "root page" (sic).
2008-08-15T23:21:16  <CIA-54> Alexander Schremmer <alex AT alexanderweb DOT de> default * 3944:a86a245db9cb 1.8/MoinMoin/ (_tests/test_PageEditor.py conftest.py):
2008-08-15T23:21:16  <CIA-54> Backed out changeset 28442c70f757, it introduced the idea to create the request more
2008-08-15T23:21:16  <CIA-54> often. I could not reproduce any test failures related to the old situation.
2008-08-15T23:21:16  <CIA-54> Thus I revert this changeset again. The changes in the PageEditor tests
2008-08-15T23:21:18  <CIA-54> will be reapplied in my next commit.
2008-08-15T23:21:20  <CIA-54> Alexander Schremmer <alex AT alexanderweb DOT de> default * 3945:3a6be2089de9 1.8/ (7 files in 4 dirs): Merge with backed out changeset.
2008-08-15T23:21:23  <CIA-54> Alexander Schremmer <alex AT alexanderweb DOT de> default * 3946:54fffaea4e22 1.8/MoinMoin/_tests/test_PageEditor.py: Reapplied the changes done in 28442c70f757 to the page editor tests.
2008-08-15T23:28:57  <xorAxAx> dennda: so, the "problem" is solved now :)
2008-08-15T23:32:27  <dennda> xorAxAx: the famous "delete items problem"?
2008-08-15T23:34:11  <xorAxAx> yes
2008-08-15T23:37:48  <dennda> ok
2008-08-15T23:38:03  <dennda> just getting rid of the last 3 remaining bugs with johill :)
2008-08-15T23:54:23  <ThomasWaldmann> xorAxAx: hg update -C 4739 ; py.test MoinMoin/security/_tests/test_security.py
2008-08-15T23:54:46  <xorAxAx> hashes please
2008-08-15T23:55:14  <xorAxAx> (of the cs)
2008-08-15T23:55:21  <ThomasWaldmann> changeset:   4739:29d166208fe0
2008-08-15T23:55:36  <xorAxAx> thats in the storage repo?
2008-08-15T23:55:42  <ThomasWaldmann> yes
2008-08-15T23:56:12  <CIA-54> Christopher Denter <moin GUESSWHAT the DASH space DASH station ROUNDTHING com> default * 4823:b8f3194fab6f 1.8-storage-cdenter/MoinMoin/storage/backends/wrapper.py: storage: make wrappers iteritems efficient
2008-08-15T23:56:13  <CIA-54> Christopher Denter <moin GUESSWHAT the DASH space DASH station ROUNDTHING com> default * 4824:29c5ab0a4950 1.8-storage-cdenter/ (3 files in 2 dirs): storage: make failing test_package tests work
2008-08-15T23:57:17  <xorAxAx> ThomasWaldmann: well, that exc is likely gone now because dennda removed all acl checking code
2008-08-15T23:57:34  <xorAxAx> so its not a state issue but a code issue :)
2008-08-15T23:58:39  <xorAxAx> dennda: ?
2008-08-15T23:58:52  <xorAxAx> dennda: "make work" != "skip"
2008-08-15T23:58:57  <xorAxAx> also you skipped the whole test ...
2008-08-15T23:59:08  <xorAxAx> you could just skip the code that requires underlay
2008-08-15T23:59:42  <ThomasWaldmann> changeset:   4743:984139f28deb
2008-08-15T23:59:55  <ThomasWaldmann> then try same with that changeset